Introducing ProLE : The Programmable Lithography Engine

Transcription

1 Introducing ProLE : The Programmable Lithography Engine Petersen Advanced Lithography, Inc Hymeadow Drive, Suite Austin, TX P: F: Info@advlitho.com

2 Outline Introduction What is ProLE TM? ProLE Components ProLE Workbench ProLE Limited Edition (ProLE-LE) ProLE Hardware ProLE Services Examples and Use Cases Lithography optimization RET validation, optimization, rule generation EDA true design for yield Copyright 2003, Petersen Page 2

3 What is ProLE? ProLE is a design-for-yield optimization solution that: Reduces litho-related systematic design failures Improves price:performance potential Reduces learning cycles before getting to market Fewer patterning related mask/design iterations Reduced wafer costs to achieve yield entitlement Improved engineering effectiveness Provides greater revenue potential thru shorter time to market Improves average selling price (ASP) and market size Provides competitive advantage by early insertion Improves performance and quality ProLE is a software-service-hardware solution using single and distributive computing capability. Copyright 2003, Petersen Page 3

4 ProLE System Overview ProLE TM Applications & Deployment Services ProLE Workbench Apps Lithography Simulation Engines 1 N ProLE Workbench Defines simulation work Generates, submits and monitors jobs Gathers and analyzes results ProLE Litho Simulation "Engines" Operate independently, in-parallel Execute jobs automatically ProLE Deployment Services Configure, calibrate & maintain litho simulation models Provide technical consulting Customize applications ProLE Hardware ProLE server and workstations Engine blade-computers in "grid" Infrastructure HW & SW 4 Copyright 2003, Petersen Page 4

5 What is the ProLE Workbench? SERIOUS Lithography Simulation Capabilities Front-load simulation setup Perform Monte Carlo Simulations Investigate Higher Order Aberrations Eliminate unnecessary simulation conditions Distribute PROLITH Simulations across a Network Grid (Cluster Computer) Tasks that are orders of magnitude too complex otherwise become manageable with the ProLE Workbench Copyright 2003, Petersen Page 5

6 The ProLE Workbench Software Drive & distribute complex PROLITH* simulation jobs Easily generate matrix-diagonals of PROLITH inputs Advanced aberration package including the ability to do aberrations of 136 Zernike terms (allows study of localized flare). Automate the output of results for Excel or ProDATA analysis Find dose-to-size and then seamlessly run properly bracketed focus-exposure matrices Quickly sort and categorize focus-exposure simulation results prior to final analysis (optional) Automatically build case tables of batch results Get multiple metrology sample cut data from multiple simulation windows and from multiple mask input files Monte Carlo variation of any PROLITH input *PROLITH and PRODATA are from KLA-Tencor Inc. Copyright 2003, Petersen Page 6

7 ProLE Workbench Workbench embeds ProLE, PROLITH, Data sorter and Automated ProData plus other software utilities. Open *.pl2 Close PROLITH & Job Builder Info Launch ProLE Client Launch ProData Display Parameters Display Help PROLITH Analyze ProLE Simulation Copyright 2003, Petersen Page 7

8 ProLE Workbench Select any PROLITH input parameter including File Based inputs Copyright 2003, Petersen Page 8

9 Select File-based PROLITH Inputs Select inputs defined by PROLITH database files and ProLE Workbench will generate simulations varying the selected files automatically. Aberration Files -.ZRN Mask Files -.MSK 1D Grayscale Masks -.GRY Source Shape Files -.SRC Spectrum Files -.ILL Available File-based Inputs Vibration Files -.VIB Resist Files -.RES Temp.(Bake)Profiles -.TPR Pupil Filter Files -.FIL CODE-V Aberrations -.INT New file type: High Order Zernikes -.HOZ User Defined Distribution -.UDD Copyright 2003, Petersen Page 9

10 Control the PROLITH Simulation Matrix Eliminate unnecessary simulations by taking control of the Simulation Matrix Use ProLE to simulate coupled inputs such as Contact Hole Width/Height, Alt. PSM Chrome Widths, and more Copyright 2003, Petersen Page 10

11 Simulate Higher Order Aberrations with PROLITH Investigate Zernike aberrations up to Z136 Correlate PROLITH aberrations with CODE-V Lens information Load and combine.zrn,.int and the new.hoz files Z1 Z22 Z44 Z72 Z129 Copyright 2003, Petersen Page 11

12 Monte Carlo Simulations with PROLITH Select up to 20 different numerical inputs Create purely random conditions, Gaussian weighted conditions or user defined distributions Note: This feature requires the ProLE cluster hardware system to execute Copyright 2003, Petersen Page 12

13 Example: Complex Aberrations Zernike Terms: Z8 (X- Coma), Z11 (45Deg. Astigmatism), Z24 (Spherical), Z99 (X- Trifoil) Conditions: -0.04, 0.07, 0.03, 0.05 Conditions: 0.15, 0.07, -0.13, Conditions: -0.03, -0.04, 0.11, 0.03 Note: Above values are in Waves Copyright 2003, Petersen Page 13

14 ProLE Limited-Edition: An Introductory Product ProLE-LE is the ProLE Workbench for a single computer Does not distribute jobs over multiple processors Does not perform Monte Carlo simulations All other features are available Huge productivity gain over ProBatch* coding, using same hardware *ProBatch is a set of commands for driving PROLITH, from KLA-Tencor Inc. Copyright 2003, Petersen Page 14

15 The ProLE Hardware ProLE Engine uses a system of compact blade computers Engine blades: 2.6GHz P4/1GB SDRAM /40GB HD ProLE server(s) Hardware infrastructure NAS Smart switches Racks Cabling UPS and clean power Software infrastructure Deployment Management Monitoring Diagnostic ProLE has a special grid-based licensing agreement for PROLITH Favorable pricing based on ProLE license management system Built in redundancy and temporary expansion Copyright 2003, Petersen Page 15

16 The ProLE Hardware PAL Cluster History Phase 0: 5 Engines 2000 Phase 1: 13 Engines 2001 Phase 2: 16 Engines 2002 Phase 3: 128 Engines 2003 Phase 4: 256 Engines 2004 Phase 5: 1024 Engines 2005 First 64 Engines of the Phase 3 expansion Copyright 2003, Petersen Page 16

17 The ProLE Hardware 1 Total Cycle Time Estimate (days) Increasing Complexity nm Node Number of ProLE TM Clients D-FE simulations Sim-volume = 9 X nm Node 90nm Node 65nm Node GHz-CPU hours 45nm Node Runtime (days) Copyright 2003, Petersen Page 17

18 ProLE Services ProLE On-site Solution Distributive computing solution Installed ProLE system at customer site, including calibration Entry level costs comparable to an EDA design seat Model calibration and monitoring, updated RET models ProLE training and software tools for customer engineers ProLE Litho-Simulation Foundry (Off-site) Solution ProLE cluster resides at PAL service center Secure on demand service Parameter generation, calibration and maintenance by PAL ProLE Consulting Service PAL analyzes design using model based OPC software and validated with ProLE at PAL service center Model calibration and monitoring Lowest cost entry point (evaluation level service) Copyright 2003, Petersen Page 18

19 ProLE Litho-Simulator Foundry Service The ProLE foundry service provides access to a larger cluster, via timesharing This allows access to much higher throughput potential ProLE timeshare Customer pays for a system that resides at PAL Several pricing options available Access is by terminal service to a secure ProLE Workbench PAL calibration services are strongly recommended Copyright 2003, Petersen Page 19

20 ProLE Consulting Service ProLE off-site consulting services: PAL analyzes design using model based OPC software and customer defined locations validated with ProLE at PAL service center Good for checking legacy OPC software PAL provides optimized critical area designs Bit-cells, NAND gates, leaf cells, etc. Physical chemical analysis of resists Lowest cost entry point (evaluation level service) ProLE on-site services: Model parameter determination, calibration and monitoring Litho-process audit Copyright 2003, Petersen Page 20

21 Who Needs ProLE? ProLE provides the power to solve problems of incredible complexity ProLE can be used by all lithographic and design-for-yield disciplines to do studies such as: Lithography optimization RET validation, optimization, rule generation EDA true design-for-yield Copyright 2003, Petersen Page 21

22 Quotes PAL s designs of a 180nm embedded SRAM IP yielded 60%. This yield is 3X the competitors; further, ProLE helped pull the product release in three months and eliminated the need for doing additional tapeouts. Mark Craig, TestChip By using ProLE, PAL helped design accurate CPL test structures quickly. In the past without ProLE, designs involved a lot of brute force optimization that took many hours. With ProLE, we are able to generate accurate results and speed up our learning. Therefore, ASML was able to deliver something to the market much sooner. Robert Socha, ASML During a beta-test of ProLE, I pulled my 100nm contact hole design project in by over 9 months. Also under a JDP with ASML, ProLE produced working chromeless phase-shift lithography patterns for our 65nm technology node process that showed production worthy imaging processes! Will Conley, Motorola Copyright 2003, Petersen Page 22

23 ProLE for Lithographers Process Development High-NA and super-high NA (immersion) RET solutions for across pitch across feature type Analysis and diagnosis of illuminator source shapes Full EMF solutions mask design Studying and designing resist formulation Process Optimization Process sensitivity using Monte Carlo methods Simulating advanced imaging system abnormalities such as aberration of 136 Zernike polynomial terms, pellicle degradation and mask blank variation Metrology Applications Line-edge-roughness for resist formulation and process optimization Two- and three-dimensional structure optimization for scatterometric and alignment target processes Copyright 2003, Petersen Page 23

24 Hot Plate Thermal Cycle Feature Accumulated Error CD 2 = 2 CD Rs ( Rs) + ( PC) + ( TC) + ( Ps) + ( Ts) CD CD Rs PC CD CD RS Ts 2 CD PC 2 2 CD TC CD CD Rs TC ( Rs)( PC) + ( Rs)( TC) + ( RS)( Ps) CD CD PC TC 2 CD Ps CD CD RS Ps CD CD PC Ps 2 CD Ts ( RS)( Ts) + ( PC)( TC) + ( PC)( Ps) CD CD CD CD CD CD + PC Ts TC Ps TC Ts where Rs = RiseTime;PC = PEBTemp; TC = TransitionTemp;Ps = PEBTime; Ts = Transition time 2 ( PC)( Ts) + ( TC)( Ps) + ( TC)( Ts) 2 2 ProLE Simulation conditions: 248nm Resist Quasar OAI; 0.80 NA; Binary mask; Results Analyzed in JMP using stepwise regression and scaling and then exported to Excel Copyright 2003, Petersen Page 24

25 Line-Edge-Roughness for 90nm 1:1 Line:Space RMS=15.8nm Approximately 60,000 sims in 30 minutes! Simulated with ProLE by varying develop and thermodynamic properties using a Monte Carlo technique. Copyright 2003, Petersen Page 25

26 Resolution Enhancement RET design optimization tools: Used for sorting and optimizing imaging technologies Weak vs. strong PSM two-beam imaging optimization Used for OPC model based rule generation, OPC optimization and validation. Use ProLE 2 to do this work efficiently or supercharge it using ProLE distributive computing solution Use Monte Carlo techniques and ProLE-IIS to find low level systematic defects 26 Copyright 2003, Petersen Page 26

27 Shortcomings of Today s Optical Proximity Correction Solutions Commercial OPC solutions: Models derived from small test bed (last generation) Models comprehend fraction of product-like features Litho tool-specific aberrations, high NA & vector affects ignored Mask errors & process biases (etch, resist) ignored Process integration effects ignored (cumulative errors) Critical for sub-0.18um DFM considerations Example feature test bed for commercial OPC/RET: 27 Copyright 2003, Petersen Page 27

28 Isofocal Region Dependence on Base Diffusivity Isofocal Region Base Diffusivity (nm 2 /s) 90nm_540nm_Pitch_+30nm_bias 90nm_180nm Pitch_+19nm_bias 540nm_Pitch_Conjugate 180nm_Pitch_Conjugate ln(acid Diffusivity) E activation for Acid and Base Diffusivity=22 Depth of Focus for 10% EL Target 90nm Pitch 180nm 540nm nm 560nm nm 620nm OPC decisions that do not comprehend the resist and etch will lead to costly mistakes! Copyright 2003, Petersen Page 28

29 ProLE RET/EDA Application: Customer Success Story Zero Yield experienced in embedded memory IP Customer requirement of optimized FEOL layers (diffusion, poly ) Focus-exposure process window optimization generated by ProLE Active, Poly, contact and metal_1 optimized to minimize systematic failures per layer and layer-to-layer Manufacturable design created in 3 days using ProLE Results 60% yield on 1st silicon with PAL-based OPC, controls yielded zero with no OPC and 20% with non-pal OPC 3 month schedule acceleration to volume production Significant process window enhancement (100% increase in DOF) No mask revisions (typical 2 or 3 spins for embedded SRAM IP) Copyright 2003, Petersen Page 29

30 ProLE RET Application: Issues Uncovered with Original Poly Design Using ProLE, PAL uncovered problems with original design manufacturability: Line CDs Space CDs Not Manufacturable! No DOF! Copyright 2003, Petersen Page 30

31 ProLE RET Application: Manufacturable Design Achieved by PAL PAL delivered a manufacturable design in 3 days! Manufacturable 0.8µm 10% EL! Copyright 2003, Petersen Page 31

32 ProLE RET/EDA Application: Critical layerto-layer imaging strategies to maximize device yield and performance Uncorrected Active/Corrected Poly Low Yield and Poor Performance Corrected Active/Corrected Poly Good Yield and Enhanced Performance A F/E = -0.2 um / 38 mj F/E = -0.2 um / 38 mj Craig, Mark J., Petersen, John S., Lund, Joshua, Gerold, David J., Chen, Nien-Po, Design, Process Integration, and Characterization for Microelectronics, Proc. SPIE Vol. 4692, p (2002). Copyright 2003, Petersen Page 32

33 ProLE EDA Application: Mask GDSII Process Simulation Silicon GDSII out Parasitic Extraction or ProDATA mask resist etch GDSII layout Parasitic Extraction Tool or ProDATA Process simulation GDSII layout Copyright 2003, Petersen Page 33

34 Lithography Drives Yield PAL is the lithography expert We embed this experience into our products Contact us to do the same for your products! Thank You µm J. V. Beach, J. S. Petersen, M. J. Maslow, D. J. Gerold, D. McCafferty, Evaluation of SCAA Mask Technology as a Pathway to the 65 nm Node, SPIE paper , µm 0.0 µm 0.1 µm 75 nm 1:1 dense lines imaged with SCAA Mask and 0.75 NA/193 nm/0.15 σ 0.2 µm Copyright 2003, Petersen Page 34